A Tailor's Server Answered in 0.47 Seconds. Google's Own Test Needed 12.3 to Paint the Page. Here Is the 20-Minute Check.

A tailor is the last person in this economy who still measures things properly.

You do not guess an inseam. You do not eyeball a shoulder and hope. You put the tape on it, you write the number down, and the number is either right or the jacket comes back. Somerset has shops on County Street doing this the same careful way they have done it for decades, and the reason those shops survive while the mall alterations counter does not is that the measurement is real.

Which is why it is worth saying plainly: your website is probably the only thing in your business you have never actually measured.

Not "does it look okay on my phone." Measured. With a number, taken the same way Google takes it, compared against a published threshold.

This article gives away the whole procedure. It takes about twenty minutes, it costs nothing, and you do not need to know a single line of code. I do this for a living and I am handing you the method, because half of you will do it yourselves and be better off, and the other half will read it, recognise the problem, and be a much better client for understanding what you are buying.

Why this matters more here than the national averages suggest#

Across four categories of Somerset small business I have audited this year, 27 sites returned a measurable mobile Largest Contentful Paint, which is the seconds-until-you-see-something number this article is about.

That median is not a national statistic about e-commerce giants. That is Somerset. Day cares, hair salons, cleaning services, a driving school. Businesses that look exactly like yours, with websites that look exactly like yours, and a customer standing in a parking lot on a three-bar connection deciding whether to wait.

What you need#

Nothing. A phone or laptop, a browser, and twenty minutes. No account, no credit card, no software to install.

Step 1 (2 minutes) — Find the URL Google actually sends people to#

Search your own shop name plus "Somerset MA" on your phone. Do not type your website address from memory. Look at what Google puts in front of a customer, and click the link on your Business Profile.

Copy that address out of the browser bar. That exact string, including whether it has www in front and whether it ends in a slash, is what you are going to test. Testing a different address than the one customers land on is the single most common way people measure the wrong thing and go away reassured.

Step 2 (3 minutes) — Run it through PageSpeed Insights#

Go to pagespeed.web.dev. It is Google's own tool, it is free, and it is the same engine that produces the numbers Google itself uses.

Paste your URL. Press Analyze. Wait. It takes thirty to ninety seconds and it will look like nothing is happening.

When it finishes, make sure you are on the Mobile tab, not Desktop. It usually opens on Mobile, but check. Your customers are on phones. A desktop score is a comforting number about a situation that is not happening.

Step 3 (5 minutes) — Read the right number, which is not the big circle#

This is the step everyone gets wrong, so read it twice.

At the top of the report you may see a section headed "Discover what your real users are experiencing." If it is there, that is field data: measurements taken from actual Chrome users who actually visited your site, on their actual phones and actual networks. Google collects it quietly in the background.

Underneath, there is a section headed "Diagnose performance issues" with a big coloured score out of 100. That is lab data: a simulation run just now, on a pretend mid-range phone, on a pretend slow network.

Field data wins every single time it exists. It is what happened. The lab score is a stress test.

Within the field section, find Largest Contentful Paint (LCP). That is the number. It answers one question: how many seconds until the biggest thing on the screen, usually your hero photograph, actually appeared? Google's threshold is 2.5 seconds or less, measured at the 75th percentile, which is Google's way of saying "three out of four visits must clear it, not the average visit."

If your site is small, the field section will be missing entirely and PageSpeed will say there is not enough data. That is not a failure. It means too few people visit for Google's dataset to have a reading, and you fall back to the lab number, knowing it runs deliberately pessimistic.

Performance budgets are not just thresholds. Much like a financial budget, they’re something consciously spent. View them as a currency to spend and trade on user-experience.

That is the mental shift that makes this whole exercise useful. You are not chasing a perfect score. You are deciding what your seconds are worth, and what you are willing to spend them on. A tailor understands this instinctively: every extra flourish costs fabric and hours, and you charge accordingly. A slideshow on your homepage costs seconds. Decide whether it is worth them.

Step 4 (2 minutes) — Know the number that is lying to you#

There is a number that every cheap "website speed checker" reports as your load time, and it is not your load time. It is server response time, sometimes called TTFB. It measures how long your host took to hand over the first file. It says nothing about when the customer saw anything.

Here is how badly the two come apart. From my own Somerset audit data:

• One Somerset site's server handed over its HTML in 0.25 seconds. Google's lab test then needed 11.4 seconds to paint the main content, and that site had no real-visitor data to soften the verdict. The server was quick. Everything after the server was not.
• Another Somerset site's server took 1.46 seconds, and its lab LCP came back at a catastrophic 40.5 seconds. Its real visitors saw content in 0.9 seconds, because real visitors arrive with things cached, on real networks, on real phones.

Same two numbers, opposite conclusions. If a vendor sends you a report that says "load time: 0.4s" and nothing else, they measured the easy thing and told you the flattering half.

Note the direction that error runs in both cases. Server response can flatter a bad site and it can also slander a good one. It is simply the wrong instrument, the way a bathroom scale is the wrong instrument for a hem.

Step 5 (5 minutes) — Read the fix list, and ignore most of it#

Scroll to Diagnostics and Opportunities. PageSpeed will list a dozen things, each with an estimated saving in seconds.

Ignore all but the top two. The list is sorted by impact, and the tail of it is noise that will make you feel hopeless for no return.

Write down the top two and their estimated savings. That is your entire to-do list. On one of the shops I measured this morning, the top two were "reduce unused JavaScript" (1.82 seconds, 305 KB) and "reduce unused CSS" (0.76 seconds, 129 KB). Two items. Two and a half seconds sitting there.

Step 6 (3 minutes to decide, 15 to do) — The free fix, yours today#

On most small business sites I look at, the Largest Contentful Paint element is the big photograph at the top of the homepage. PageSpeed will actually tell you which element it picked, in the LCP section of the report. And on a lot of those sites, that photograph is a straight-off-the-camera file being shrunk down by the browser after it has already been downloaded in full.

You are making a customer download a 4,000-pixel-wide image so their phone can display it 400 pixels wide. It is the digital equivalent of shipping a bolt of fabric to alter a cuff.

Here is the fix, free, no code:

1. On your homepage, right-click the big top image and choose "Open image in new tab." Look at the address. If it ends in .jpg or .png, you have room to improve it.
2. Go to squoosh.app. It is free, made by Google's Chrome team, and it runs entirely inside your browser. Nothing uploads anywhere.
3. Drag your image in. On the right-hand panel, set the format to WebP. Under Resize, set the width to 1600 pixels.
4. Watch the file size counter at the bottom. In my experience a 3 MB camera photo usually lands somewhere near 200 KB at these settings. Your file will differ. The counter tells you the truth for your image, so use that rather than my estimate.
5. Download it and swap it into your website wherever you normally edit your homepage. In Wix, Squarespace, WordPress or a UENI-style builder, that is the same "replace image" button you already know.
6. Wait ten minutes, then run Step 2 again and compare.

That is the whole thing. Fifteen minutes and no invoice. I am not going to promise you a specific number of seconds back, because I have not measured your site. What I can promise is that Step 2 will tell you exactly what it was worth, which is the entire point of learning to measure.

What I actually found this morning#

I ran this exact procedure on three Massachusetts tailoring and alterations shops on the morning of 25 August 2026, found by hand from public listings. Two of the three list a Somerset location. All three are anonymised here, because none of them asked to be graded in public and it would be poor manners to do it anyway.

Three shops, three completely different stories, and not one of them is the story you would guess from looking at the sites.

Site A looks like the loser on server response and is the strongest of the three. Its host took 3.44 seconds to answer me from a cold connection, which sounds dreadful. But it has enough real traffic that Google holds field data on it, and real visitors see the main image in 2.6 seconds, a whisker outside the threshold, with 74% of visits already landing in the good band.

And now I have to hold myself to my own standard, because I just did the thing I warned you about. I quoted you the flattering half. Site A's full field record is graded SLOW overall by Google, because its real-visitor server response sits at 2.7 seconds at the 75th percentile, which is genuinely slow, and its First Contentful Paint is 2.7 seconds too. So: the picture arrives acceptably fast, and the wait before anything at all appears is the weak point. That is a hosting and server conversation, not a redesign. Separately, the site carries no LocalBusiness structured data at all, which is a local search problem rather than a speed one, and cheaper to fix than either.

Site B is the trap. Fast server, small page at 63 KB, structured data present, and a lab LCP of 12.3 seconds. Every naive speed checker on the internet would give this site a gold star. Its top two opportunities were unused JavaScript and unused CSS, together about 434 KB of code that nothing on the page uses, most likely dragged in by a page builder's default theme. There is no field data for it, so I cannot tell you what real visitors experience, only that Google's simulated mid-range phone struggled badly.

Site C is the one that keeps me up at night. A real shop with a real listing, and a website address on that listing that did not respond when I tried it this morning. I am not going to publish a number I did not measure, so: not measured. But a listing that links to a broken site converts worse than a listing with no link at all, because the customer has already formed an opinion before they ever reach you.

Where the twenty minutes actually goes#

Do it once a quarter. Put it in the same drawer as the fire extinguisher check.

What I could not measure, and will not pretend otherwise#

My business-discovery lookup returned an error this morning, so this is not a census of Somerset tailors. It is three shops I found by hand and measured properly. The Somerset median of 6.1 seconds comes from four other categories audited earlier this year, not from tailors. Where a site had no field data I have said so rather than dressing a lab estimate up as what real customers experience. And I have not measured anyone's Google Business Profile, review recency, or social posting frequency here, because that was not what this walkthrough was for.

When to stop and call someone#

Run the six steps. If your field LCP comes back under 2.5 seconds, you are done, close the tab and go back to work. If it comes back between 2.5 and 4, do the image fix and re-measure; there is a decent chance that closes it.

If it comes back over 4 seconds after you have shrunk the hero image, the problem is structural: theme bloat, a page builder loading code you never use, a host that is oversubscribed. That is when it is worth paying someone, and now you can say exactly what you want done and check afterwards whether it happened. You have the before number. Insist on the after number.

I am in Fall River and I work with shops across Somerset, Swansea and Westport. If you run the check and want a second pair of eyes on the result, send me the PageSpeed link and I will tell you honestly whether it is worth spending money on. Sometimes the answer is no.

774.559.8992 · Joshua.Amado@AvalonPartner.com